冬眠。引用 SINGLE_TABLE 继承类的集合字段在 fething 时忽略 DiscriminatorValue

Question

问题：一对多集合字段忽略鉴别器列并获取所有条目而不是专用子类。

我有带有继承策略 SINGLE_TABLE 的基超类。

@DiscriminatorColumn(name = "state")
@Data
@Table(name = "named_entity", indexes = @Index(columnList = "state"))
@Inheritance(strategy = InheritanceType.SINGLE_TABLE)
@Entity
public class BaseNamedEntity {
    @Id
    private UUID id = UUID.randomUUID();
    @Column(insertable = false, updatable = false)
    @Setter(AccessLevel.PROTECTED)
    @Enumerated(EnumType.STRING)
    private EntityState state; // declared for jpql purposes
    private String name;
    @ManyToOne 
    @JoinColumn(name = "collection_holder_id", referencedColumnName = "id")
    private CollectionHolderEntity otmCollection; // field will be important later
}

状态“活动”和“已删除”的两个子类：

@Entity
@DiscriminatorValue("active")
public class ActiveNamedEntity extends BaseNamedEntity {
}
@Entity
@DiscriminatorValue("deleted")
public class DeletedNamedEntity extends BaseNamedEntity {
}

基本功能（如 spring 数据 CRUD 存储库）按预期工作：

((CrudRepository<ActiveNamedEntity, UUID>) activeDao).existsById(UUID.randomUUID());
// -> Hibernate: select count(*) as col_0_0_ from named_entity activename0_ where activename0_.state='active' and activename0_.id=?
// Discriminator column is checked as expected

我尝试引入与子类之一保持一对多关系的实体：

@Entity
@Table(name = "collection_holder")
@Data
public class CollectionHolderEntity {
    @Id
    private UUID id = UUID.randomUUID();
    @EqualsAndHashCode.Exclude
    @ToString.Exclude
    @OneToMany(mappedBy = "otmCollection", fetch = FetchType.EAGER) // Eager to avoid transactions in tests
    private List<ActiveNamedEntity> oneToManyCollection = new ArrayList<>(); // ActiveNamedEntity is subclass with corelated state = "active"
}

以及一旦尝试获取集合字段，生成的语句似乎会忽略鉴别器列：

collectionHolderDao.findById(collectionHolderEntity.getId()) // reloading entity
                .orElseGet(() -> Assertions.fail("Expected to find holder " + collectionHolderEntity));
// -> Hibernate: select collection0_.id as id1_0_0_, onetomanyc1_.collection_holder_id as collecti4_1_1_, onetomanyc1_.id as id2_1_1_, onetomanyc1_.id as id2_1_2_, onetomanyc1_.name as name3_1_2_, onetomanyc1_.collection_holder_id as collecti4_1_2_, onetomanyc1_.state as state1_1_2_ from collection_holder collection0_ left outer join named_entity onetomanyc1_ on collection0_.id=onetomanyc1_.collection_holder_id where collection0_.id=?
// join contains only id reference, but not state check:
// left outer join named_entity onetomanyc1_ on collection0_.id=onetomanyc1_.collection_holder_id

其中将请求预期行为，其中包含连接的状态检查：左外连接named_entity onetomanyc1_ on collection0_.id=onetomanyc1_.collection_holder_id 和 onetomanyc1_.state="active"

我是否缺少规范的某些部分？

也许我应该在某个地方明确指定它应该向生成的请求引入鉴别器列。如果是这样 - 预计如何处理多对多关系？

到目前为止，我尝试：

1. 将拥有子类关系的字段（本例中的ActiveNamedEntity）移动，但似乎没有任何变化。
2. 切换获取类型也得到相同的结果

在这个例子中我使用了org.springframework.boot:spring-boot-starter-data-jpa:2.4.6 with 与 org.hibernate:hibernate-core:5.4.31.Final 相关联

Answer 1

我仍然不认为这是预期的行为，但这里有一些东西可以让它开箱即用：

1. 添加@org.hibernate.annotations.Where(clause)对于具有相关查询的每个子类：

@Entity
@DiscriminatorValue("active")
@Where(clause = "state='active'")
public class ActiveNamedEntity extends BaseNamedEntity { }

@Entity
@DiscriminatorValue("deleted")
@Where(clause = "state='deleted'")
public class DeletedNamedEntity extends BaseNamedEntity { }

或者直接对于集合属性：

/*...*/
    @Where(clause = "state='active'")
    private List<ActiveNamedEntity> oneToManyCollection = new ArrayList<>();

第一个似乎更可取，因为它将普遍工作。

2. 添加@org.hibernate.annotations.Filter（和相应的filterDef）并为每个会话启用它。更多内容请参见此讨论。
   我对这种方法没有任何运气，但理论上它应该有效。